A RAG-powered internal research assistant for CollabCircle.

Orbit is a specialized AI chatbot designed to serve as the central knowledge hub for CollabCircle. Powered by Google's Gemini API and LangChain, Orbit ingests internal research papers, policy documents, and meeting minutes to provide accurate, context-aware answers to organization members.

• Walled Garden: Answers questions only based on the internal CollabCircle documents you upload.
• Zero-Cost Embeddings: Uses a local embedding model (HuggingFace) to process documents without hitting API rate limits.
• Privacy First: Uses a local vector database (ChromaDB) to store document knowledge securely.
• Smart Fallback: Automatically handles deployment environments (Linux/Cloud) vs. local development (Windows).

• Frontend: Streamlit
• LLM: Google Gemini Flash (via gemini-flash-latest for stability)
• Embeddings: HuggingFace (all-MiniLM-L6-v2 running locally)
• Vector Store: ChromaDB
• Orchestration: LangChain

• Python 3.10+
• A Google Cloud API Key (for Gemini)

1. Clone the repository:

   git clone [https://github.com/smri29/Orbit.git](https://github.com/smri29/Orbit.git)
   cd Orbit

2. Create a virtual environment:

   python -m venv .venv
   # Windows:
   .venv\Scripts\activate
   # Mac/Linux:
   source .venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

Create a .env file in the root directory and add your Google API key:

GOOGLE_API_KEY="AIzaSy....."

streamlit run app.py

Note: The first time you run the app, it will download the embedding model (~90MB). This is normal.

5. Reboot the app. Orbit handles the SQLite database requirements automatically via pysqlite3-binary.

Internal Tool for CollabCircle.